a model of environmental factors driving fish …pubs.iclarm.net/resource_centre/wf_1001.pdf ·...

Baran E., Makin I., Baird I.G. 2003 BayFish: a model of environmental factors driving fish production in the Lower Mekong Basin. Contribution to the Second International Symposium on Large Rivers for Fisheries. Phnom Penh, Cambodia, 11-14 February 2003.

BAYFISH: A MODEL OF ENVIRONMENTAL FACTORS DRIVING FISH PRODUCTION

IN THE LOWER MEKONG BASIN

Eric Baran, Ian Makin1, Ian G. Baird2

WorldFish Center (formerly ICLARM) P.O. box 500 GPO, 10670 Penang, Malaysia

ABSTRACT In tropical floodplain systems local populations are generally highly dependent upon the system’s natural aquatic resources. In such systems the annual fish production depends on a combination of biological and physical parameters, principally 1) hydrological factors; 2) environmental factors and 3) fish migrations. Developing management plans is complicated as assessing the role of each factor is usually made difficult by their diversity, their interactions or feed-back loops, and by the frequent absence of data on certain factors. We present here a tool developed to overcome these difficulties with the aim of facilitating the management of large tropical rivers. Despite the absence of adequate regional or national statistics and comprehensive time series data, a review of existing studies has allowed, the identification of ten factors driving the fish production in the Mekong River Basin. They are 1) hydrological factors: water level, flood duration; flood timing, flooding regularity; 2) floodplain factors: bank types, flooded zone land cover, dry season refuges, bank types, turbidity; 3) biological factors: longitudinal or lateral migrations. This assessment provided the framework for the production of a model integrating the driving environmental parameters and their interactions. Fishing, that also influences fish production, is not yet part of this model focussing on hydrological and ecological factors. The model is based on Bayesian networks. The above variables are interconnected in a logical fashion and related to fish production with the connections being expressed in terms of probabilities. The model calculates conditional probabilities at each level and the overall trend resulting from the sum of interactions within the system. When quantitative information is not available, probabilities are based on local knowledge drawn from experienced biologists and fishers. This model was developed for three groups of fishes (black fishes, white fishes and opportunists) and three geographic sectors (Upper Mekong, Tonle Sap system and Mekong Delta). The natural production that can be expected for each fish group in each sector is qualitatively expressed by a percentage between “Bad” and “Good”. This result could have been converted into tonnes of fish had statistical time series been available. The model is transparent and user-friendly. Variables having the strongest influence on fish production are identified and each variable can be easily varied to assess the consequences of various management options on fish production. Corresponding author: E. Baran WorldFish Center, P.O. box 500 GPO, 10670 Penang, Malaysia. Email: [email protected] Telephone: (604) 6261606 Fax: (604) 6265530

Running title: BayFish: a Mekong fish production model

Keywords: Fish production; Environment; Floodplains; Mekong; Modelling; Bayesian networks

1 International Water Management Institute, PO box 2075, Colombo, Sri Lanka. E-mail: [email protected] 2 Global Association for People and the Environment, P.O box 860, Pakse, Lao PDR. E-mail: [email protected]

1


INTRODUCTION The environment The Mekong River flows through China, Burma (Myanmar), the Lao PDR, Thailand, Cambodia and Vietnam. Among the large rivers of the world, the Mekong ranks tenth in length and third in mean maximum discharge (Welcomme 1985). This results in large monsoon-driven floods, and floodplains are a dominant feature of the Lower Basin, their total surface amounting to 84,000 km2, the surface of Ireland (Scott 1989, Lacoursiere et al. 1998). The extent of these floodplains and wetland favours the abundance and diversity of fish; the total number of species is estimated at 1200 (Rainboth 1996) and the total fish catches basinwide amounts to an estimated 1.5 million tonnes (Sverdrup-Jensen 2002). Another feature of the Mekong River is the importance of fish migrations over long distances, in particular between countries (Lagler 1976, Pantúlu 1986, Roberts & Warren 1994, Poulsen et al. 2000). Numerous species migrate between spawning grounds (in Laos or Northern Cambodia) and feeding grounds in the wetlands and floodplains of Southern Laos, Cambodia and Vietnam. The use of these inundated zones by fishes has long been identified as a factor which strongly influences their total production (Petillot 1911, Chevey & Le Poulain 1940, Baird and Phylavanh 1999). Importance of the fish resource Rice and fish from capture fisheries are the two major staple food resources of the 60 million people living in the basin. Fish from capture fisheries (75% of the total yield, Sverdrup-Jensen 2002), supplemented by reservoir fisheries, aquaculture and by aquatic animals such as frogs and snails (Shoemaker et al. 2001, Balzer et al 2002), constitute the dominant source of protein in the region. The populations of the countries of the Lower Mekong Basin have one of the highest rates of fish consumption in the world with an average of 33.4 kg/person/year (Baran & Baird 2001), compared to the world average that is around 16 kg/person/year (FAO 2000). For instance, in Cambodia, fish constitute up to 65-75 per cent of total protein in the diet (Guttman 1999). Access to this common property resource is particularly important for rural poor and landless people. However, at the national and regional levels the importance of water-dependant resources for food security and livelihoods has so far been poorly recognised (Jensen 2001, IIRR et al. 2001). The need for management tools In relation to its recent economic expansion, the Mekong River Basin has been experiencing a dramatic increase in the demand for water, arising from irrigated agriculture, hydropower production, and domestic and industrial uses. The optimised and sustainable use of the water and related resources of the Basin is dependant upon allocation rules to be defined among the riparian countries. This requires hydrological and water quality monitoring, development of modelling tools, simulation of the effects of basin development scenarios and the definition of rules for water allocation “so that water flow and ecological systems are maintained while Basin resources are developed” (MRC 2000). The importance of aquatic resources in the diet and livelihood of rural communities, in particular poor people, requires that fish production be taken into consideration in management plans and strategies. Furthermore the dependence of aquatic resources upon flooding and floodplains requires that land management and environmental issues be also factored in to avoid development options that threaten food security in parts of the basin. In such a context the usually distinct objectives of modelling are needed simultaneously: identification of critical questions; building of a conceptual framework; synthesis of existing knowledge and identification of gaps; production of simulation scenarios for managers. In 2000-2001, ICLARM-The WorldFish Center, in collaboration with the Mekong River Commission and the International Water Management Institute (IWMI), developed a hydro-biological research program on the relationships between hydrology, environmental factors and fish production in the Mekong River Basin. Research activities have continued in 2002 in order to integrate interactions between hydrological, biological and environmental factors into a computer model (Baran & Cain 2001, Baran 2002). The resulting model, named BayFish, indicates the probability of having a good or bad natural fish production in the Mekong Basin, as a function of hydrological and environmental parameters. Production is undestood here literally as the biomass produced in a given year, and not as the yield resulting from a fishing effort. The model aims to assess the consequences of various management options regarding flows and floodplain environment on natural fish production basinwide, and to provide scenarios for planning and decision making at the basin level.

2


MATERIAL AND METHODS Constraints As in several developing countries, the absence of data in the region is a major impediment to modelling. Classical fisheries models require data on harvested biomass but also, depending upon the models, on stock virgin biomass, growth, length frequencies, natural mortality, recruitment, or fishing effort and selectivity. In the case of the Mekong River Basin and given the history of the region, reliable data of this nature are not available even for dominant commercial fish species. Despite some recent efforts to improve representativeness, “the total reported production from inland waters appears to be under-estimated by a factor of between at least 2.5 and 3.6” in national catch statistics (Coates 2002). Furthermore most fisheries models (with the exception of Halls et al. 2001) do not address floodplain environmental variability and do not allow for the simulation of different options for land and water use. Alternatively models based on an ecosystem approach such as Ecopath or Ecosim (Christensen & Pauly 1992, 1993, Walters et al. 1997) require data on biomass of certain animal and vegetal taxa (e.g. macrophytes, zooplankton, meiobenthos, etc…) and data on trophic flows; both of which are non-existent for the Mekong system. Use of existing knowledge as an alternative approach Despite the paucity of quantitative data and the absence of time-series basinwide, the ecological knowledge of Mekong fishing communities is rich (Sokheng et al. 1999, Poulsen et al. 2000, MRC 2001, Baird 2002) and constitutes a source of integrative and low-cost information appropriate for exploratory studies (Poizat & Baran 1997). A number of ecological descriptions of the Mekong system and fisheries (e.g.: Chevey & Le Poulain 1940, Lagler 1976) provide useful sources of information to be considered. The BayFish model has been developed with the aim of overcoming the paucity of quantitative data and of integrating this qualitative bio-ecological information. Bayesian networks have been demonstrated to be a convenient tool for the modelling of knowledge (Cowell et al. 1999), and BayFish was developed following this approach. Model conceptual framework The conceptual framework presented here is based on a thorough analysis of hydrological, environmental and ecological factors influencing the natural fish production basinwide. The steps followed are detailed below. • In order to assess the variability of the water level and the extent of each type of vegetation flooded, a hydrological model was developed by IWMI (Kite 2000, 2001). This model provides, on a daily basis, the discharge and river water levels basinwide. The availability of a digital terrain model and vegetation maps for the Tonle Sap sub-basin allowed the calculation of the area of each type of vegetation flooded, on the same temporal basis. • The analysis of the monitoring of the Cambodian bag net fishery, done in collaboration with the project “Management of the Cambodian freshwater Capture Fisheries” (MRC/DoF/DANIDA), showed the strong correlation between water level and opportunistic fish species catches in Cambodia (Deap Loeung 1999, Baran et al. 2001). • Analysis of fish monitoring data from Southern Lao PDR allowed migration-driven abundance patterns to be identified for 110 Mekong species (Baran & Baird 2002). It also showed the role and importance of dry season deep-water refuges for fish in the mainstream Mekong River. • Other parameters important to fish have been identified from literature review and from interviews, and have been compiled (Baran and Coates 2000, Baran et al. 2001). Ten factors have been identified as driving natural fish production in the Mekong River Basin. They are 1) hydrological factors (water level, flood duration; flood timing, flooding regularity); 2) floodplain factors (bank type, type of vegetation flooded, zone land cover, presence of dry season refuges, turbidity); 3) biological factors (possibility for fish to undertake longitudinal or lateral migrations). These variables have been linked in the following conceptual framework (Figure 1).

3


TOTAL NATURAL FISH PRODUCTION

Water level

Flooded area

Flood duration

Flooding regularity

Flood timing

HYDROLOGICAL FACTORS MIGRATIONS

Type of vegetation

flooded

Dry season refuges

Bank type

ENVIRONMENTAL FACTORS

Turbidity



Water level

Flooded area

Flood duration

Flooding regularity

Flood timing

HYDROLOGICAL FACTORS MIGRATIONS

Type of vegetation

flooded

Dry season refuges

Bank type

ENVIRONMENTAL FACTORS

Turbidity

Figure 1: Parameters driving the fish production basinwide

In addition to these natural factors, it should be noted that fishing is a factor also having a strong influence on fish production or productivity, on fish sizes and on species assemblages. (Laë 1995, van Zalinge et al. 2000, Baran et al. 2003). Due to the complexity of fisheries in the Mekong Basin, to the absence of reliable catch and effort statistics basinwide and to our initial focus on environmental management, fishing has not been included in the model at this stage. Bayesian modelling Bayesian networks are based on three basic components (Jensen 1996):

- nodes representing system variables (e.g. flood duration; type of vegetation flooded,). Each variable can be continuous or discrete with a finite set of mutually exclusive states (e.h. long/short; forest/shrubs/grass). Driving or independent variables are called parent nodes and driven or dependant variables are called child nodes. - links representing causal relationships between these nodes (from parent to child nodes, i.e. from cause to effect); - probabilities attached to combination of states, and quantifying the believed relationships between connected nodes.

• For each driving variable (“parent nodes”) the states are defined after the consultation of experts, and a probability is set for each state; this is the first step of the parametrization process (“elicitation of probabilities”). For example in the Tonle Sap system,

- there is a 80% chance that the bank type is a natural floodplain (i.e.it is estimated that in 80% of cases the banks are in relation with a non-altered floodplain); - there is a 10% chance that the bank type is a man-made embankment (i.e. it is considered that 10% of the Great Lake perimeter is restrained by National roads 5 and 6); - there is a 10% chance that the bank type is a valley (i.e. it is estimated that 10% of the banks are bordered by hills -“phnoms”- and thus constrained from flooding by a natural feature, as in a valley).

• For each variable driven by independent variables, experts attribute a probability of occurrence to every combination of states in the parent nodes; this is the second and last step of the parametrization process. For instance in the Tonle Sap system

- there is a 100% chance that “flooded area” is “large” when “water level” is “high” and “bank type” is “natural plain”, but - there is only a 30% chance that “flooded area” is “large” when “water level” is “low” and “bank type” is “natural plain”.

• For all child nodes subject to the combined influence of two or more parent nodes, the software calculates the probability of each state that results from the combination of states of parent nodes. In other words "child probabilities" resulting from several "parent probabilities" are calculated after the Bayes formula for combined probabilities (hence the name of Bayesian network), down to the last child node (the probability of a good natural fish production basinwide). This represents the overall trend resulting from the sum of interactions within the

4


system. Figure 2 gives an example of this process for two parent variables and one child variable in the case of the Tonle Sap system.

Flooded areaLargeSmall

61.838.2

Child node

In the Tonle Sap system- in 80% of cases the bank type is a natural floodplain- in 10% of cases the bank type is an embankment- in 10% of cases the bank type is a valley(Parameters set by the user)

In the Tonle Sap system and given the climate of these recent years- there is a 60% chance that the water level is high- there is a 40% chance that the water level is low(Probabilities set by the user)

Bank TypeNatural plainEmbankmentValley

80.010.010.0

Parent node

Water levelHighLow

60.040.0

Parent node

In the Tonle Sap system, given the probabilities defined for "Water level" and "Bank type", the model computes that there is a 61.8% chances that the overall flooded area is "Large" and a 38.2% chance that the overall flooded area is "Small"

SmallLargeBank typeWater level

8020ValleyLow9010EmbankmentLow7030Natural plainLow7030ValleyHigh8020EmbankmentHigh0100Natural plainHigh

Chances of having a flooded areaStatus of variables

The user sets the probabilities for a combination of status; e.g.: - it is very likely (100% chance) to have a large flooded area when the water level is high and the bank type is a natural plain- it is likely (80% chance) to have a small flooded area when the water level is low and the bank type is a valley

Calculation of combined probabilities(Bayes formula)


61.838.2

Child node

In the Tonle Sap system- in 80% of cases the bank type is a natural floodplain- in 10% of cases the bank type is an embankment- in 10% of cases the bank type is a valley(Parameters set by the user)

In the Tonle Sap system and given the climate of these recent years- there is a 60% chance that the water level is high- there is a 40% chance that the water level is low(Probabilities set by the user)

Bank TypeNatural plainEmbankmentValley

80.010.010.0

Parent node

Water levelHighLow

60.040.0

Parent node

In the Tonle Sap system, given the probabilities defined for "Water level" and "Bank type", the model computes that there is a 61.8% chances that the overall flooded area is "Large" and a 38.2% chance that the overall flooded area is "Small"

SmallLargeBank typeWater level

8020ValleyLow9010EmbankmentLow7030Natural plainLow7030ValleyHigh8020EmbankmentHigh0100Natural plainHigh

Chances of having a flooded areaStatus of variables

The user sets the probabilities for a combination of status; e.g.: - it is very likely (100% chance) to have a large flooded area when the water level is high and the bank type is a natural plain- it is likely (80% chance) to have a small flooded area when the water level is low and the bank type is a valley

Calculation of combined probabilities(Bayes formula)

Figure 2: Example of interactions between 3 nodes of a Bayesian network (example taken from BayFish / Tonle Sap system, Cambodia)

By modifying the probabilities in parent nodes one at a time, the user can identify the variables of the system and observe the result of changed parameters on the intermediate and final child nodes. Different scenarios can thus be considered, and sensitivity analysis can point out variables that are critically important in the general behaviour of the system. BayFish specificities • BayFish was developed using Netica software (Norsys Corporation); this software is simple, user-friendly and a user-version of the operating software can be downloaded freely on Internet at www.norsys.com. • BayFish was developed for three distinct groups of fishes (Welcomme 1985, Hoggarth et al. 1999): - “white fish”(they undertake mostly long longitudinal migrations and move back to the mainstream of perennial rivers when water recedes) - “black fishes (they undertake mostly short lateral migrations and move back to ponds when the water recedes) - opportunists (small cyprinids of short life span able to reproduce within one year; a dominant and apparently growing part of catches in the Mekong Basin (Van Zalinge & Thuok 1999, Van Zalinge et al. 2000). The importance of these fishes in catches in Cambodia justified the creation of a special group for them. • BayFish was also developed for three distinct zones in the Mekong Basin: - upland zone (Northern highlands and Khorat plateau, Lao PDR; valley-dominated bank type; migrations of fishes to central zone) - central zone (Lower Mekong plain, including the Tonle Sap river and lake; large non-modified floodplain) - delta (from below Phnom Penh to the sea; floodplain controlled by embankments). • The initial parametrization of the model was done by the authors, then modified by four Mekong River Commission fisheries and environment specialists (see acknowledgements) until a consensus was reached on the final set of parameters. Each geographical zone was parameterized independently according to the local environment and hydrology. The detail of the parameterization (i.e. probabilities used) is given in annex I.

5


RESULTS Three sub-models and an integrative one. For each geographical zone (upper catchment, Tonle Sap, Delta), the interaction between hydro-environmental variables and their resulting impact on the natural production of three fish groups (white fish, black fish, opportunists) has been be modeled. In each zone the fish production of each group is expressed by a probability between 0 and 1 (i.e. between “Bad” and “Good”). Figure 3 details this model for the Tonle Sap sub-system.

Flooding regularityRegularIrregular

60.040.0

Flood durationLongMediumShort

20.040.040.0

Flood timingEarlyMidLate

20.060.020.0


61.838.2

Flood heightHighLow

60.040.0

Bank TypeNat PlainEmbankmentValley

80.010.010.0

HYDROLOGYGoodNeutralPoor

42.619.737.7

ENVIR. for White fishesGoodPoor

60.439.6

ENVIR. for Black fishesGoodPoor

48.851.2

RefugesAvailableNone

50.050.0

Flood zone land coverGrass riceShrubForest

30.026.044.0

BLACK FISH PRODUCTIONGoodNormalBad

33.125.841.1

WHITE FISH PRODUCTIONGoodNormalBad

29.632.537.9

OPPORTUNISTS PRODUC.GoodPoor

68.231.8

Flooding regularityRegularIrregular

60.040.0

Flood durationLongMediumShort

20.040.040.0

Flood timingEarlyMidLate

20.060.020.0


61.838.2

Flood heightHighLow

60.040.0

Bank TypeNat PlainEmbankmentValley

80.010.010.0

HYDROLOGYGoodNeutralPoor

42.619.737.7

ENVIR. for White fishesGoodPoor

60.439.6

ENVIR. for Black fishesGoodPoor

48.851.2

RefugesAvailableNone

50.050.0

Flood zone land coverGrass riceShrubForest

30.026.044.0

BLACK FISH PRODUCTIONGoodNormalBad

33.125.841.1

WHITE FISH PRODUCTIONGoodNormalBad

29.632.537.9

OPPORTUNISTS PRODUC.GoodPoor

68.231.8

Figure 3: BayFish model of fish production for the Tonle Sap sub-system

The three zones have been connected by links representing migrations and spatial interactions; their respective natural fish productions are combined to give the total natural fish production basinwide. Outputs of the model for the natural fish production of the three geographical zones are given in Figure 4.

TS+∆ Prod. OpportunistsGoodBad

66.733.3

UC Production White fishGoodNormalBad

21.733.245.0

UC Production Black fishGoodNormalBad

31.026.642.4

Upper catchment

TS Production Black fishGoodNormalBad

33.125.841.1

TS Production White fishGoodNormalBad

25.732.641.7

Tonle Sap

Delta Production White FishGoodNormalBad

14.532.552.9

Delta Production Black fishGoodNormalBad

38.625.635.8

Delta


66.733.3


66.733.3


21.733.245.0


31.026.642.4

Upper catchment


21.733.245.0


31.026.642.4


21.733.245.0


31.026.642.4

Upper catchment


33.125.841.1


25.732.641.7

Tonle Sap


33.125.841.1


25.732.641.7


33.125.841.1


25.732.641.7

Tonle Sap


14.532.552.9


38.625.635.8

Delta


14.532.552.9


38.625.635.8


14.532.552.9


38.625.635.8

Delta

Figure 4: Outputs of the model for the three fish groups of the three geographical zones

The output of the overall model is therefore a probability of having a good fish production basinwide, depending upon the production in each zone. The overall model is represented (without attached probabilities, for readibility) in Figure 5.

6


Upper catchment

Tonle Sap

Delta

FISH PRODUCTION

TS Envir for W hite Fish

TS Flooding regularity

TS Flooded area

TS Bank Type

TS Production W hite fish

TS Envir for Black Fish

Delta Flood tim ing

TLS Flood timing

TS Flood duration

TS W ater level

UC Hydrological factors

UC Flooded area

UC Production Black fish

UC Production W hite f ish

UC Flood duration

TS Vegetation cover TS Refuges

TS Envir. for Opportunists

UC Envir. for W hite f ish

UC TurbidityUC RefugesUC Vegetation floodedUC Bank Type

UC Flooding regularity

UC Flood timing

UC W ater level

UC Envir. for Black fish

Delta Flooding regularity

Delta Flooded area

Delta Bank Type

Delta Flood duration

Delta Flood height Delta Refuges Delta TurbidityDelta Vegetation flooded

Delta Envir for W hite fish

Delta Envir for Black fish

Delta Production W hite Fish

Delta Production Black fish

Delta Envir. for Opportunists

Delta Hydrological factors

TS Hydrological factors

TS Production Black fish

White fish

Bad Good

F IS H PRO DUCT IO NGoodBad

58.441.6

O pportun is ts

Bad Good

Black fish

Bad Y. Good .

Upper catchment

Tonle Sap

Delta

FISH PRODUCTION

TS Envir for W hite Fish

TS Flooding regularity

TS Flooded area

TS Bank Type

TS Production W hite fish

TS Envir for Black Fish

Delta Flood tim ing

TLS Flood timing

TS Flood duration

TS W ater level

UC Hydrological factors

UC Flooded area

UC Production Black fish

UC Production W hite f ish

UC Flood duration

TS Vegetation cover TS Refuges

TS Envir. for Opportunists

UC Envir. for W hite f ish

UC TurbidityUC RefugesUC Vegetation floodedUC Bank Type

UC Flooding regularity

UC Flood timing

UC W ater level

UC Envir. for Black fish

Delta Flooding regularity

Delta Flooded area

Delta Bank Type

Delta Flood duration

Delta Flood height Delta Refuges Delta TurbidityDelta Vegetation flooded

Delta Envir for W hite fish

Delta Envir for Black fish

Delta Production W hite Fish

Delta Production Black fish

Delta Envir. for Opportunists

Delta Hydrological factors

TS Hydrological factors

TS Production Black fish

White fish

Bad Good

F IS H PRO DUCT IO NGoodBad

58.441.6

O pportun is ts

Bad Good

Black fish

Bad Y. Good .

Figure 5: Summary of the Mekong BayFish model

Sensitivity analysis With a classical statistical approach, testing the interactions among 23 variables would require more than a century-long time-series. In the Bayesian approach, a sensitivity analysis allows for the identification of the variables that contribute most to the variability of a given node. This consists in a systematic variation of each probability of each network node; the output probabilities are investigated, some variables having a considerable effect on the final node, while others hardly reveal any influence (Coupé et al. 2000). In practice the percentage which each parent node (i.e. driving variable) contributes to the total entropy reduction in the system (i.e. the reduction of uncertainty) was calculated and plotted. This approach can be used in the knowledge acquisition process to focus attention on the most influential parameters, thus improving the efficiency of monitoring (Coupé et al. 1999). In the case of BayFish the sensitivity analysis outlined the critical variables of the system (Figure 6).

Tonle Sap: flooded area

Tonle Sap: presence of refuges

0

5%

10%

15%

20%

25%

Tonle Sap: flood duration

Upper Mekong: presence of refugesTonle Sap: flooded land coverDelta: flooded areaUpper Mekong: flooded area

Perc

enta

ge o

f con

tribu

tion

to to

tal e

ntro

py

Driving variables in each zone

Tonle Sap: flooded area

Tonle Sap: presence of refuges

0

5%

10%

15%

20%

25%

Tonle Sap: flood duration

Upper Mekong: presence of refugesTonle Sap: flooded land coverDelta: flooded areaUpper Mekong: flooded area

Perc

enta

ge o

f con

tribu

tion

to to

tal e

ntro

py

Driving variables in each zone Figure 6: Variables contributing most to the natural fish production.

For readability variables “Water level” and “Bank type” have been merged into variable “Flooded area”. Ordering of variables results from their contribution ot the total entropy of the system, as identified after a sensitivity analysis.

This figure shows that the area of flooding around the Tonle Sap is the most influential parameter driving the fish production; among 7 influential factors, the area flooded is present 3 times, highlighting the importance of natural

7


flooding for natural fish production. Among environmental factors, the nature of the vegetation flooded in the Tonle Sap floodplain and the availability of dry season refuges (in the Tonle Sap and upstream) also rank high in effects on overall production. DISCUSSION The strength of BayFish is that it is based on an extensive review of information on fish and fisheries in the Basin. It has been developed to overcome the paucity of reliable statistics in the Mekong Basin by integrating expert and local knowledge. In its current version all quantitative information has been coded as probabilities (“codage”), but the software permits the replacement of qualitative information by data when they become available (Figure 7).

Bayesian model

Quantitative data

Qualitative information

Equations

Deduced trends Codage

Codage

Bayesian model

Quantitative data

Qualitative information

Equations

Deduced trends Codage

Codage

Figure 7: Possible inputs for a Bayesian network. In grey inputs used for the BayFish model.

The model is intuitive and user-friendly. Sources are open, size is small (60 kB). Experience has shown that such a model is easily understood and generates interest and discussions among users. However BayFish is not a dynamic model, in the sense that it does not integrate a temporal component. This reflects the fact that information backing this model is integrative and does not rely upon regular monitoring of the system modelled. It is possible for Bayesian networks to integrate temporal evolutions by connecting several models, each of them representing a certain time period, and parameters being re-examined from one step to the other according to the outputs of the previous model. This is a possible development of BayFish, as long as it can rely on studies of trends in Mekong Basin land and water use. BayFish is hampered, as other models would be, by i) the absence of data systematically gathered at the basin level, in particular for fisheries; ii) the limited information on fish productivity by habitat type and habitat distributions in floodplains; and iii) the unknown fishing effort basin wide. Subsequently the production is expressed in terms of probability, but the catch, depending upon the fishing effort, cannot be quantified/forecasted. This would have been done had reliable statistical time series been available. The issue of fishing effort is important, as the size of fishes and the species composition of catches are known to be largely influenced by the fishing pressure (Welcomme 1995). Integrating fishers and their fishing effort in the next versions of this model are planned developments. Among other perspectives are the integration of i) floodplain inundation patterns and ii) available habitats at different water levels for the rest of the Basin. Last, it can be argued that BayFish only reflects the knowledge and beliefs of those who took part in its parameterization (Bayesian networks being also called belief networks). In the context of management of tropical environments with little scientific data, this can be considered as a strength. Bayesian networks indeed allow for the integration of experiences drawn by specialised resource persons, and permit the synthesis of the best available information about a complex system. The model itself integrates the information entered at different levels and provides a synthesis that is beyond the reach of individual experts. This is shown for instance in the sensitivity analysis, which provides a simple ranking of the most influential parameters when interactions among them are integrated. However the way the locally available information is integrated into the model is an important issue. In Bayfish the parametrization was made by six experts in Mekong fisheries and environment. However one might recommend a larger consultation of experts, in particular of fishers (Cain 2001). This is of particular importance in future versions of this model integrating the effect of fishing on the fish resource. A greater number of contributors would then require the definition of a formalized consultation process for a proper (consensual, averaged or weighted) parametrization of the model.

8


CONCLUSIONS BayFish as it stands is mostly a tool for discussions about development options among and between regional agencies, national management bodies and companies conducting Environmental Impact Assessments. Bayesian networks have been developed in the mid-90’s to build Decision Support Systems. Being intuitive and easy to compute or modify, they constitute a good tool for building a common representation of a system between stakeholders willing to discuss consequences of various management scenarios. They have been developed in recent years for natural resources management (review in Cain 2001), fisheries management (Varis & Kuikka 1997, Kuikka et al. 1999, Marcot et al. 2001) and watershed management (Borsuk et al. 2001). In the same spirit, multi-agent systems (Weiss 1999, Ferber 1999) have also been used to holistically model floodplain fisheries and integrated river basin management (Bousquet 1994, FIRMA 2000). Tharme (2002) has shown that in the field of river management a new generation of holistic models was mostly being designed and applied in tropical countries. This probably reflects the fact that in the tropics people are in more direct relation to natural environments and their dynamics, and dependant upon rivers and their aquatic resources. Beyond refinements of the current model, the long term perspective is to develop this approach as part of an integrated modelling package to assist in floodplains management, in particular in the tropics. This would correspond to a decision-support system encompassing three successive levels: fish and their environment; then users of the environmental resources; and then the local, national and regional management bodies having a stake in a river basin. In order to model these three layers of increasing complexity, variables and their interactions have to be identified and qualified or quantified, which requires extensive interdisciplinary interactions and consultations. Last, unlike BayFish that focusses on fish only, and thus has a simple perspective, an integrative decision-support system will be faced with a diversity of perspectives from a variety of stakeholders. This issue of integrating perspectives and achieving a consensus among stakeholders through modelling is an active field of research (e.g. Moss 2000, Borsuk et al. 2001, Trebuil et al. 2002) and a challenge for water and floodplain management in the Mekong Basin. ACKNOWLEDGEMENTS The authors would like to acknowledge the contribution of Nicolaas van Zalinge, Anders Poulsen, Heng Kong and Deap Loeung to the parametrization and refinements of this model. David Coates, John Valbo Jorgensen, Tyson Roberts, Pierre Dubeau and Terry Warren have also contributed by their suggestions to the initial conceptual framework. Thanks also to Gayathri Jayasinghe for comments on the statistical aspects of the model.

9


BIBLIOGRAPHY Baird I. G. (2002) Some comments on conducting assessments of fish and fisheries based on local ecological knowledge. In: Haggen, N. B. Neis, and I.G. Baird (eds.) Putting fishers knowledge to work. Blackwell Sciences. In press. Baird I. G. & Phylavanh B. (1999) Fishes and Forests: fish floods and the importance of seasonally flooded riverine habitats for Mekong River fish. Environmental Protection and Community Development in Siphandone Wetlands Project, CESVI, Pakse, Lao PDR. 46 pp. Balzer T., Balzer P. & Pon S. (2002) Traditional use and availability of aquatic biodiversity in rice-based ecosystems. I. Kampong Thom Province, Kingdom of Cambodia. FAO/Netherlands Partnership Programme "Awareness of Agricultural Biodiversity". Fao, Rome. 16 pp + annexes. Baran E. (2002) Modelling the Mekong fisheries: what can be done? In: Sverdrup-Jensen S. (ed.) Fisheries in the Lower Mekong Basin: status and perspectives. MRC Technical Paper nº 6. Mekong River Commission, Phnom Penh, Cambodia.pp 70-73. Baran E. & Baird I. (2001) Approaches and tools for sustainable management of fish resources in the Mekong River Basin. Contribution to the International Symposium on Biodiversity Management and Sustainable Development in the Lancang-Mekong River Basin. 4-7 December 2001, Xishuangbanna, Yunnan, China. Baran E. & Baird I.G. (2003) Khone Falls Artisanal fisheries (Mekong River, Southern Laos): analysis of a monitoring database. Part I: Data preparation, Fisheries description, Comparison of two fisheries, Migrations, Deep pools. WorldFish Center working paper. In press. Baran E. & Cain J. (2001) Ecological approaches of flood-fish relationships modelling in the Mekong River. Proceedings of the National workshop on Ecological and Environmental Modelling, 3-4 September 2001, Universiti Sains Malaysia, Penang, Malaysia. In press. Baran E. & Coates D. (2000) Hydro-biological models for water management in the Mekong River. In: Al-Soufi (Ed.) Proceedings of the Workshop on hydrologic and environmental modelling in Mekong Basin, Phnom Penh 11-12 September 2000. Mekong River Commission publications, Pp. 328-334. Baran E., Van Zalinge N. & Ngor Peng Bun (2001) Analysis of the Cambodian bagnet ("dai") fishery data. ICLARM, Penang, Malaysia, Mekong River Commission Secretariat and Department of Fisheries, Phnom Penh, Cambodia. 62 pp. Baran E., Van Zalinge N. & Ngor Peng Bun (2003) Floods, floodplains and fish production in the Mekong Basin: present and past trends. Proceedings of the 2001 Asian Wetlands Symposium, 27-30 August 2001, Penang, Malaysia. In press. Baran E., Van Zalinge N., Ngor Peng Bun, Baird I. G. & Coates D. (2001) Fish resource and hydrobiological modelling approaches in the Mekong Basin. ICLARM, Penang, Malaysia and the Mekong River Commission Secretariat, Phnom Penh, Cambodia. Borsuk M. E., Clemen R. T., Maguire L. A. & K. H. Reckhow (2001) Stakeholder values and scientific modeling in the Neuse River watershed. Group Decision and Negotiation 10, 355-373. Bousquet F. (1994). Distributed artificial intelligence and object-oriented modelling of a fishery. Mathematical Computer Modelling 2018, 97-107. Cain J. (2001) Planning improvements in natural resources management. Guidelines for using Bayesian networks to support the planning and management of development programmes in the water sector and beyond. Centre for Ecology & Hydrology (CEH) Wallingford, UK. 124 pp. Chevey P. & Le Poulain F. (1940) La pêche dans les eaux douces du Cambodge 5e mémoire. Travaux de l'Institut Océanographique de l'Indochine. 195 pp. Christensen V. & Pauly D. (eds.) (1993) Trophic models of aquatic ecosystems. ICLARM Conf. Proc. 26, 390 p. Christensen V. & Pauly D. 1992 Ecopath II- a system for balancing steady-state ecosystem models and calculating network characteristics. Ecol. Modelling 61, 169-185 Coates D. (2002) Inland capture fishery statistics of Southeast Asia: Current status and information needs. Asia-Pacific Fishery Commission, Bangkok, Thailand. FAO/RAP Publication No. 2002/11, 113 p. Coupé V.M.H., Peek N.B., Ottenkamp J. & Habbema J.D.F. (1999) Using sensitivity analysis for efficient quantification of a belief network. Artificial Intelligence in Medicine, 17, 223-247.

10


Coupé V.M.H., van der Gaag L.C. & Habbema J.D.F. (2000) Sensitivity analysis: an aid for belief-network quantification. Knowledge Engineering Review, 15, 1-18. Cowell R.G., Dawid A.P., Lauritzen S.L. & Spiegelhalter D.J. (1999) Probabilistic networks and expert systems. Springer-Verlag, New-York. 370 pp. Deap Loeung (1999) The bagnet (Dai) fishery in the Tonle Sap River. Proceedings of the Annual meeting of the Department of Fisheries of the Ministry of Agriculture, Foresty and Fisheries, 19-21 January 1999. MRC/Danida project for management of the freshwater capture fisheries of Cambodia, Phnom Penh, Cambodia.Van Zalinge, Thuok & Loeung Eds. 141-149 FAO (2000) The state of world fisheries and aquaculture. Editorial group FAO information division 142 p. Ferber J. (1999) Multi-agent systems: an introduction to distributed artificial intelligence. Reading, MA, Addison-Wesley. FIRMA (2000) Freshwater Integrated Resource Management with Agents. A project supported by European Union (“Framework 5 Programme for Research and Development”) and by the European Commission (“Key Action on Sustainable Management and Quality of Water programme”). Web site: http://firma.cfpm.org Guttman H. (1999) Rice fields fisheries - a resource for Cambodia. NAGA 22 (2) 11-15 Halls A.S., Kirkwood G.P. & Payne A.I. (2001) A dynamic pool model for floodplain-river fisheries. Ecohydrology & Hydrobiology 1 (3) 323-339 Hoggarth D.D., Cowan V.J., Halls A.S., Aeron-Thomas M., McGregor J A., Garaway C.A., Payne A.I. & Welcomme R.L. (1999) Management guidelines for Asian floodplain river fisheries. Part 1: A spatial, hierarchical and integrated strategy for adaptive co-management. FAO fisheries technical paper 384/1 63 pp. IIRR, IDRC, NACA and ICLARM (2001) Utilizing different aquatic resources for livelihoods in Asia: a resource book. International Institute for Rural Reconstruction, International Development Research Center, Food and Agriculture Organization of the United Nations, Network of Aquaculture Centers in Asia-Pacific, International Center for Living Aquatic Resources Management. 416 pp. Jensen F.V. (1996) An introduction to Bayesian networks. UCL press, London. 178 pp. Jensen J.G. (2001) Managing fish, flood plains and food security in the lower Mekong basin. Water Science and Technology 43 (9) 157-164 Kite G. (2000) Developing a hydrological model for the Mekong Basin. Impacts of basin development on Fisheries productivity. IWMI working paper n° 2, 141 pp. Kite G. (2001) Modelling the Mekong: hydrological simulation for environmental impact studies. Journal of Hydrology (Amsterdam) 1/4, 253; 1-13 Kuikka, S., Hildén, M., Gislason, H., Hansson, S., Sparholt, H. & Varis, O. (1999) Modeling environmentally driven uncertainties in Baltic cod (Gadus morhua) management by Bayesian influence diagrams. Canadian journal of fisheries and aquatic sciences 56 (4) 629-641. Laë R. (1995) Climatic and anthropogenic effects on fish diversity and fish yields in the central delta of the Niger river. Aquatic Living Resources 8 43-58 Lagler K. F. (1976) Fisheries and integrated Mekong river basin development. The University of Michigan, School of natural resources - Executive volume 363 p. Marcot, B. G., R. S. Holthausen, M. G. Raphael, M. Rowland, & M. Wisdom (2001) Using Bayesian belief networks to evaluate fish and wildlife population viability under land management alternatives from an environmental impact statement. Forest Ecology and Management 153 (1-3) 29-42. Moss S., Downing T.E. & Rouchier J. (2000) Demonstrating the role of stakeholder participation: an agent based social simulation model of water demand policy response. Centre for policy modelling - Manchester Metropolitan University Business school - report No 76. MRC (2000) Mekong River Commission annual report. Mekong River Commission, Phnom Penh. 33 p. MRC (2001) Local knowledge in the study of river fish biology: experiences from the Mekong Mekong Development series No.1 22 pp. + 5 ann. Pantúlu V. R. (1986) Fish of the Lower Mekong Basin. In: Davies B. R., & Walker K. F. (Eds.) The ecology of River systems. Junk Publishers, Dordrecht, The Nederlands.

11


Petillot L. (1911) La pêche et les poissons. Une richesse du Cambodge - Augustin Challamel Ed, librairie maritime et coloniale. 169 p. Poizat G. & Baran E. (1997) Fishermen's knowledge as background information in tropical fish ecology: a quantitative comparison with fish sampling results. Environmental Biology of Fishes, 50 435-449. Poulsen A.F. et al. (2000) Fish migrations and spawning habits in the Mekong mainstream - A survey using local knowledge. Assessment of Mekong Fisheries: Fish Migrations and Spawning and the Impact of Water Management Project (AMFC) Vientiane, Lao P.D.R. February 2000. 132 pp. Roberts T. R. & Warren T. J. (1994) Observations on fishes and fisheries in Southern Laos and Northeastern Cambodia, October 1993-February 1994. Nat. Hist. Bull. Siam Soc. 42 87-115 Sokheng C. et al. (1999) Fish migrations and spawning habits in the Mekong mainsteam - a survey using local knowledge. (Basin-wide) MRC - Assessment of Mekong fisheries - AMFP Report 2/99, November 1999 Vientiane. 56 p. Sverdrup-Jensen S. (ed.) (2002) Fisheries in the Lower Mekong Basin: status and perspectives. MRC Technical paper nº 6, Mekong River Commission, Phnom Penh. 103 pp. Tharme R. E. (2002) A global perspective on environmental flow assessment: emerging trends in the development and application of environmental flow methodologies for rivers. Submitted to Rivers Research and Applications. Trebuil G., Bousquet F., Baron C. & Shinawatra-Ekasingh B. (2002) Collective creation of artificial worlds can help concrete natural resource management problems: a Northern Thailand experience. Proceedings of the International symposium-Sustaining food security and managing natural resources in southeast Asia- Challenges for the 21st Century - January 8-11, 2002 at Chiang Mai, Thailand. Van Zalinge N. P. & Thuok N. (1999) Present status of Cambodia's freshwater capture fisheries and management implications. Proceedings of the Annual meeting of the Department of Fisheries of the Ministry of Agriculture, Foresty and Fisheries, 19-21 January 1999. MRC/Danida project for management of the freshwater capture fisheries of Cambodia, Phnom Penh, Cambodia.Van Zalinge, Thuok & Loeung Eds. 11-20 Van Zalinge N., Nao Thuok, Touch Seang Tana & Deap Loeung (2000) Where there is water, there is fish? Cambodian fisheries issues in a Mekong River Basin perspective. In: M. Ahmed and P. Hirsch (eds.) Common property in the Mekong: issues of sustainability and subsistence. ICLARM Studies and Reviews 26, p. 37-48. Varis, O., S. Kuikka (1997) Joint use of multiple environmental assessment models by a Bayesian meta-model: the Baltic salmon case. Ecological modelling 102 (2-3): 341-351. Walters C.J., Christensen V. & Pauly D. (1997) Structuring dynamic models of exploited ecosystems from trophic mass-balance assessments. Rev. Fish Biol Fish. 7 137-172 Weiss, G. (ed.) (1999). Multiagent systems : a modern approach to Distributed Artificial Intelligence, MIT Press. Welcomme R. L. (1985) River fisheries. FAO Fisheries Technical Paper 262 330 p. Welcomme R.L. (1995) Relationships between fisheries and the integrity of river systems. Regulated rivers: research and management 11 121-136.

12


ANNEX I: VARIABLES AND PARAMETERS OF THE BAYFISH MODEL

Each number is a probability entered after consultations with local experts. The tables below should be read as follows: Upper catchment: Flood duration

Long Short 0.5 0.5

Tonle Sap: Vegetation flooded Tonle Sap Water level

Grass or rice

Shrub Forest

High 0.4 0.2 0.4 Low 0.15 0.35 0.5

In the Tonle Sap system, - when the water level is high, there is a 40% chance that the floodedvegetation is rice, 20% chances that it is shrubs and 40% chances that it is aforest - when the water level is low, there is a 15% chance that the flooded vegetationis rice, 35% chances that it is Shrubs and 50% chances that it is a forest

In the Upper catchment, there is a 50% chance that the flood duration isLong, and 50% chances that the flood duration is Short

Parent variables Upper catchment Upper catchment: Water level

High Low 0.6 0.4

Upper catchment: Flood duration Long Short 0.5 0.5

Upper catchment: Flood timing Early Mid Late 0.2 0.6 0.2

Upper catchment: Flooding regularity Regular Irregular

0.6 0.4 Upper catchment: Vegetation flooded

Grass or rice Shrub Forest 0.2 0.4 0.4

Upper catchment: Bank type Natural plain Embankment Valley

0.1 0.1 0.8 Upper catchment: Refuges

Available None 0.5 0.5

Upper catchment: Turbidity High Low 0.5 0.5

Tonle Sap Tonle Sap: Water level

High Low 0.6 0.4

Tonle Sap: Flood duration Long Medium Short 0.2 0.4 0.4

Tonle Sap: Flood timing Early Mid Late 0.2 0.6 0.2

Tonle Sap: Flooding regularity Regular Irregular

0.6 0.4

13


Tonle Sap: Vegetation flooded Tonle Sap Water level Grass or rice Shrub Forest High 0.4 0.2 0.4 Low 0.15 0.35 0.5 Tonle Sap: Bank type

Natural plain Embankment Valley 0.8 0.1 0.1

Tonle Sap: Refuges Available None

0.5 0.5 Delta Delta: Water level

High Low 0.6 0.4

Delta: Flood duration Long Short 0.5 0.5

Delta: Flood timing Early Mid Late 0.2 0.6 0.2

Delta: Flooding regularity Regular Irregular

0.6 0.4 Delta: Vegetation flooded

Grass or rice Shrub Forest 0.7 0.2 0.1

Delta: Bank type Natural plain Embankment Valley

0.7 0.25 0.05 Delta: Refuges

Available None 0.5 0.5

Delta: Turbidity High Low 0.5 0.5

Child variables Upper Catchment Upper catchment: Flooded area UC Water level UC Bank type Large Small High Natural plain 1 0 High Embankment 0.2 0.8 High Valley 0.3 0.7 Low Natural plain 0.3 0.7 Low Embankment 0.1 0.9 Low Valley 0.2 0.8

14


Upper catchment: Hydrological factors UC Flooding regularity

UC Flooded area UC Flood duration

UC Flood timing Good Neutral Poor

Regular Large Long Early 1 0 0 Regular Large Long Mid 0.95 0.05 0 Regular Large Long Late 0.7 0.2 0.1 Regular Large Short Early 0.7 0.2 0.1 Regular Large Short Mid 0.6 0.2 0.2 Regular Large Short Late 0.5 0.1 0.4 Regular Small Long Early 0.4 0.3 0.3 Regular Small Long Mid 0.3 0.3 0.4 Regular Small Long Late 0.2 0.2 0.6 Regular Small Short Early 0.2 0.2 0.6 Regular Small Short Mid 0.1 0.2 0.7 Regular Small Short Late 0.1 0.2 0.7 Irregular Large Long Early 0.8 0.1 0.1 Irregular Large Long Mid 0.7 0.2 0.1 Irregular Large Long Late 0.6 0.2 0.2 Irregular Large Short Early 0.4 0.2 0.4 Irregular Large Short Mid 0.3 0.3 0.4 Irregular Large Short Late 0.3 0.2 0.5 Irregular Small Long Early 0.3 0.3 0.4 Irregular Small Long Mid 0.2 0.2 0.6 Irregular Small Long Late 0.1 0.2 0.7 Irregular Small Short Early 0.1 0.2 0.7 Irregular Small Short Mid 0 0.1 0.9 Irregular Small Short Late 0 0 1 Upper catchment: Environment for White fish: UC Refuges UC Turbidity UC Vegetation flooded Good Poor Available High Grass or rice 0.2 0.8 Available High Shrub 0.7 0.3 Available High Forest 0.8 0.2 Available Low Grass or rice 0.3 0.7 Available Low Shrub 0.8 0.2 Available Low Forest 0.9 0.1 None High Grass or rice 0.1 0.9 None High Shrub 0.5 0.5 None High Forest 0.6 0.4 None Low Grass or rice 0.2 0.8 None Low Shrub 0.6 0.4 None Low Forest 0.7 0.3 Upper catchment: Environment for Black fish: UC Refuges UC Turbidity UC Vegetation flooded Good Poor Available High Grass or rice 0.7 0.3 Available High Shrub 0.6 0.4 Available High Forest 0.6 0.4 Available Low Grass or rice 0.9 0.1 Available Low Shrub 0.8 0.2 Available Low Forest 0.8 0.2 None High Grass or rice 0.3 0.7 None High Shrub 0.2 0.8 None High Forest 0.2 0.8 None Low Grass or rice 0.5 0.5 None Low Shrub 0.4 0.6 None Low Forest 0.4 0.6 Upper catchment: Production of Black fish Hydrological factors UC Environment for Black

fish Good Normal Bad

Good Good 1 0 0 Good Poor 0 0.4 0.6 Neutral Good 0.6 0.4 0 Neutral Poor 0.1 0.4 0.5 Poor Good 0.3 0.5 0.2 Poor Poor 0 0 1

15


Upper catchment:Production of White fish UC Hydrological factors

UC Environment for White fish

TS Production of White fish

Good Normal Bad

Good Good Good 1 0 0 Good Good Normal 0.9 0.1 0 Good Good Bad 0.8 0.1 0.1 Good Poor Good 0 0.6 0.4 Good Poor Normal 0 0.55 0.45 Good Poor Bad 0 0.45 0.55 Neutral Good Good 0.5 0.5 0 Neutral Good Normal 0.4 0.6 0 Neutral Good Bad 0.35 0.65 0 Neutral Poor Good 0.1 0.6 0.3 Neutral Poor Normal 0.05 0.6 0.35 Neutral Poor Bad 0 0.55 0.45 Poor Good Good 0 0.5 0.5 Poor Good Normal 0 0.45 0.55 Poor Good Bad 0 0.4 0.6 Poor Poor Good 0 0.1 0.9 Poor Poor Normal 0 0.05 0.95 Poor Poor Bad 0 0 1 Tonle Sap Tonle Sap: Flooded area Tonle Sap Water level Tonle Sap Bank type Large Small High Natural plain 1 0 High Embankment 0.2 0.8 High Valley 0.3 0.7 Low Natural plain 0.3 0.7 Low Embankment 0.1 0.9 Low Valley 0.2 0.8 Tonle Sap: Hydrological factors Tonle Sap Flooding regularity Tonle Sap Flooded area Tonle Sap Flood duration Tonle Sap Flood timing Good Neutral Poor Regular Large Long Early 0.4 0.3 0.3 Regular Large Long Mid 0.3 0.3 0.4 Regular Large Long Late 0.2 0.25 0.55 Regular Large Medium Early 1 0 0 Regular Large Medium Mid 0.95 0.05 0 Regular Large Medium Late 0.7 0.2 0.1 Regular Large Short Early 0.7 0.2 0.1 Regular Large Short Mid 0.6 0.2 0.2 Regular Large Short Late 0.5 0.1 0.4 Regular Small Long Early 0.3 0.2 0.5 Regular Small Long Mid 0.2 0.2 0.6 Regular Small Long Late 0.1 0.1 0.8 Regular Small Medium Early 0.4 0.3 0.3 Regular Small Medium Mid 0.3 0.3 0.4 Regular Small Medium Late 0.2 0.2 0.6 Regular Small Short Early 0.2 0.2 0.6 Regular Small Short Mid 0.1 0.2 0.7 Regular Small Short Late 0.1 0.2 0.7 Irregular Large Long Early 0.35 0.4 0.25 Irregular Large Long Mid 0.3 0.4 0.3 Irregular Large Long Late 0.25 0.35 0.4 Irregular Large Medium Early 0.8 0.1 0.1 Irregular Large Medium Mid 0.7 0.2 0.1 Irregular Large Medium Late 0.6 0.2 0.2 Irregular Large Short Early 0.4 0.2 0.4 Irregular Large Short Mid 0.3 0.3 0.4 Irregular Large Short Late 0.3 0.2 0.5 Irregular Small Long Early 0.2 0.2 0.6 Irregular Small Long Mid 0.1 0.1 0.8 Irregular Small Long Late 0.01 0.1 0.89 Irregular Small Medium Early 0.3 0.3 0.4

16


Irregular Small Medium Mid 0.2 0.2 0.6 Irregular Small Medium Late 0.1 0.2 0.7 Irregular Small Short Early 0.1 0.2 0.7 Irregular Small Short Mid 0 0.1 0.9 Irregular Small Short Late 0 0 1 Tonle Sap: Environment for White fish: TS Refuges TS Vegetation flooded Delta Refuges Upper catchment Refuges Good Poor Available Grass or rice Available Available 0.15 0.85 Available Grass or rice Available None 0.1 0.9 Available Grass or rice None Available 0.15 0.85 Available Grass or rice None None 0.1 0.9 Available Shrub Available Available 0.85 0.15 Available Shrub Available None 0.65 0.35 Available Shrub None Available 0.9 0.1 Available Shrub None None 0.6 0.4 Available Forest Available Available 0.95 0.05 Available Forest Available None 0.75 0.25 Available Forest None Available 0.85 0.15 Available Forest None None 0.7 0.3 None Grass or rice Available Available 0.2 0.8 None Grass or rice Available None 0.05 0.95 None Grass or rice None Available 0.2 0.8 None Grass or rice None None 0.05 0.95 None Shrub Available Available 0.6 0.4 None Shrub Available None 0.55 0.45 None Shrub None Available 0.5 0.5 None Shrub None None 0.45 0.55 None Forest Available Available 0.7 0.3 None Forest Available None 0.65 0.35 None Forest None Available 0.75 0.25 None Forest None None 0.5 0.5 Tonle Sap: Environment for Black fish Refuges Vegetation flooded Good Poor Available Grass or rice 0.55 0.45 Available Shrub 0.6 0.4 Available Forest 0.8 0.2 None Grass or rice 0.4 0.6 None Shrub 0.2 0.8 None Forest 0.3 0.7 Tonle Sap:Environment for Opportunists TS Vegetation flooded TS Hydrological factors Good Poor Grass or rice Good 1 0 Grass or rice Neutral 0.6 0.4 Grass or rice Poor 0.1 0.9 Shrub Good 0.9 0.1 Shrub Neutral 0.75 0.25 Shrub Poor 0.6 0.4 Forest Good 0.8 0.2 Forest Neutral 0.7 0.3 Forest Poor 0.55 0.45

17


Tonle Sap: Production of Black fish Hydrological factors TS Environment for Black fish Good Normal Bad Good Good 1 0 0 Good Poor 0 0.4 0.6 Neutral Good 0.6 0.4 0 Neutral Poor 0.1 0.4 0.5 Poor Good 0.3 0.5 0.2 Poor Poor 0 0 1 Tonle Sap: Production of White fish TS Hydrological factors TS Environment for White fish Good Normal Bad Good Good 1 0 0 Good Poor 0 0.5 0.5 Neutral Good 0.4 0.6 0 Neutral Poor 0 0.6 0.4 Poor Good 0 0.5 0.5 Poor Poor 0 0 1 Delta Delta: Flooded area Delta Water level Delta Bank Large Small High Natural plain 1 0 High Embankment 0.2 0.8 High Valley 0.3 0.7 Low Natural plain 0.3 0.7 Low Embankment 0.1 0.9 Low Valley 0.2 0.8

Delta: Environment for White fish

Delta Refuges Delta Turbidity Delta Vegetation flooded

Good Poor

Available High Grass or rice 0.2 0.8 Available High Shrub 0.7 0.3 Available High Forest 0.8 0.2 Available Low Grass or rice 0.2 0.8 Available Low Shrub 0.7 0.3 Available Low Forest 0.8 0.2 None High Grass or rice 0.1 0.9 None High Shrub 0.5 0.5 None High Forest 0.6 0.4 None Low Grass or rice 0.1 0.9 None Low Shrub 0.5 0.5 None Low Forest 0.6 0.4 Delta: Environment for Black fish Delta Refuges Delta Turbidity Delta Vegetation flooded Good Poor Available High Grass or rice 0.8 0.2 Available High Shrub 0.7 0.3 Available High Forest 0.7 0.3 Available Low Grass or rice 0.8 0.2 Available Low Shrub 0.7 0.3 Available Low Forest 0.7 0.3 None High Grass or rice 0.4 0.6

High Shrub 0.3 0.7 None High Forest 0.3 0.7 None Low Grass or rice 0.4 0.6 None Low Shrub 0.3 0.7 None Low Forest 0.3 0.7

None

18


Delta: Hydrological factors Delta Flooding regularity

Delta Flooded area Delta Flooding duration Delta Flood timing Good Neutral Poor

Regular Large Long Early 1 0 0 Regular Large Long Mid 0.95 0.05 0 Regular Large Long Late 0.7 0.2 0.1 Regular Large Short Early 0.7 0.2 0.1 Regular Large Short Mid 0.6 0.2 0.2 Regular Large Short Late 0.5 0.1 0.4 Regular Small Long Early 0.4 0.3 0.3 Regular Small Long Mid 0.3 0.3 0.4 Regular Small Long Late 0.2 0.2 0.6 Regular Small Short Early 0.2 0.2 0.6 Regular Small Short Mid 0.1 0.2 0.7 Regular Small Short Late 0.1 0.2 0.7 Irregular Large Long Early 0.8 0.1 0.1 Irregular Large Long Mid 0.7 0.2 0.1 Irregular Large Long Late 0.6 0.2 0.2 Irregular Large Short Early 0.4 0.2 0.4 Irregular Large Short Mid 0.3 0.3 0.4 Irregular Large Short Late 0.3 0.2 0.5 Irregular Small Long Early 0.3 0.3 0.4 Irregular Small Long Mid 0.2 0.2 0.6 Irregular Small Long Late 0.1 0.2 0.7 Irregular Small Short Early 0.1 0.2 0.7 Irregular Small Short Mid 0 0.1 0.9 Irregular Small Short Late 0 0 1 Delta: Environment for Opportunists Delta Vegetation flooded Delta Hydrological factors for

Opportunists Good Poor

Grass or rice Good 1 0 Grass or rice Neutral 0.6 0.4 Grass or rice Poor 0.1 0.9 Shrub Good 0.9 0.1 Shrub Neutral 0.75 0.25 Shrub Poor 0.6 0.4 Forest Good 0.8 0.2 Forest Neutral 0.7 0.3 Forest Poor 0.55 0.45 Delta: Production of Black fish Delta Hydrological factors for Black fish Delta Environment for Black

fish Good Normal Bad

Good Good 1 0 0 Good Poor 0 0.4 0.6 Neutral Good 0.6 0.4 0 Neutral Poor 0.1 0.4 0.5 Poor Good 0.3 0.5 0.2 Poor Poor 0 0 1

19


Delta: Production of White fish Delta Hydrological factors for White fish Delta Environment for White

fish Tonle Sap Production of White fish

Good Normal Bad

Good Good Good 1 0 0 Good Good Normal 0.95 0.05 0 Good Good Bad 0.8 0.2 0 Good Poor Good 0.05 0.45 0.5 Good Poor Normal 0 0.5 0.5 Good Poor Bad 0 0.5 0.5 Neutral Good Good 0.5 0.5 0 Neutral Good Normal 0.4 0.6 0 Neutral Good Bad 0.3 0.7 0 Neutral Poor Good 0.05 0.55 0.4 Neutral Poor Normal 0 0.6 0.4 Neutral Poor Bad 0 0.5 0.5 Poor Good Good 0.05 0.45 0.5 Poor Good Normal 0 0.5 0.5 Poor Good Bad 0 0.5 0.5 Poor Poor Good 0 0.05 0.95 Poor Poor Normal 0 0 1 Poor Poor Bad 0 0 1

Production basinwide Basin: Opportunists Tonle Sap Environment for Opportunists Delta Environment for

Opportunists Good Bad

Good Good 1 0 Good Poor 0.7 0.3 Poor Good 0.3 0.7 Poor Poor 0 1

20

a model of environmental factors driving fish …pubs.iclarm.net/resource_centre/wf_1001.pdf ·...

Documents